For the development of an energy-efficient bipedal robot prototype, an automated parameter identification framework is presented, making use of gradient-based optimization to minimize the residual error between the simulated and measured system states and locate optimal parameters. For the identification, the prototype is mounted on a test-stand, such that its legs emulate the motion of either a single or a double pendulum, its angular displacements and velocities measured by encoders integrated within the motors at each joint, which are used as reference in the identification. The identification framework is formulated to have a global nature, allowing multiple measurements sets from either single or double pendulum configuration of the prototype to be combined in the same optimization routine, enabling parameter identifications valid over a range of operating regimes of the prototype. The identification framework is then implemented to identify the dissipation effects in the drivetrain of the prototype, which is critical for energy efficiency. To facilitate the simultaneous identification of dissipation along with other mechanical and electrical parameters of the prototype, the dissipation torque within each joint is modelled as a polynomial velocity-dependent ansatz function. Furthermore, to investigate the transient nature of the dissipation with changing joint angular velocity regimes, low-velocity and high-velocity measurements from the prototype are combined in the identification framework. For low velocities, constant and linear dissipation effects dominate in the hip and knee joints respectively, whereas for high velocities, non-linear dissipation effects, more specifically with a cubic dependence on velocity, is observed for both the joints. Above all, the observed dissipation torques are several orders of magnitude lesser than the input torques for both low and high velocity regimes, confirming the energy-efficient design of the prototype drivetrain.